In 1994, Bonetti et al 1 proposed the concept of a family of tumors comprised of angiomyolipoma, clear cell sugar tumors and lymphagioleiomyomatosis. The proposal was based on observations that these 3 lesions shared a morphologically and immunophenotypically distinctive cell type which the authors had previously designated "perivascular epithelioid cell" 23. Cells that are probably synonymous with these perivascular epithelioid cells (PECs) were first noted in renal angiomyolipomas by Apitz in 1943 4. These cells lack a normal anatomic homologue, have spindle to epithelioid shapes with clear to eosinophilic cytoplasm, display a predilection for perivascular arrangements, and display immunoreactivity for melanocytic markers such as HMB-45, micropthalmia transcription factor (miTF), Melan A, Mart-1, HMSA-1 and to a lesser extent, muscular markers such as actin and desmin 56789101112131415. PECs are envisioned by the Bonetti group as possessing a phenotypic plasticity wherein the cells may assume a spindled appearance and be more likely to be positive for muscular markers, an epithelioid appearance associated with a higher frequency of immunoreactivity for melanocytic markers, or various phenotypic modulations in between 1415. Over the subsequent one and a half decades, an increasing number of neoplasms putatively composed of PECs have been reported from a wide variety of anatomic locations under a similarly wide variety of appellations. The term PEComa was introduced in 1996 by Zamboni et al to describe one such case arising in the pancreas 16. Perivascular epithelioid cell tumors received formal recognition in 2 monographs published under the auspices of the World Health Organization (WHO) in 2002 and 2003. 1017. In the WHO classification of soft tissue neoplasms 10, PEComa was defined as "mesenchymal tumors composed of histologically and immunohistochemically distinctive perivascular epithelioid cells", including angiomyolipoma, clear cell sugar tumors, lymphagioleiomyomatosis, clear cell myomelanocytic tumor of the falciform ligament/ligament teres and other unusual clear cell tumors at various locations 10. Clear cell myomelanocytic tumor of the falciform ligament/ligament teres 18 is presently not considered sufficiently distinctive and is now generally included in the latter group of "unusual clear cell tumors" at various locations. The designation PEComa-NOS (perivascular epithelioid cell tumors-not otherwise specified) has been applied to these "unusual clear cell tumors" to avoid applying the same designation – PEComa – to the family of lesions as well as a constituent subset 5678. However, the unqualified "PEComa" designation is recognized in the WHO monographs 1017 and will be used henceforth in this article to describe the aforementioned constituent subset. PEComas display an overwhelming female preponderance and appear to be anatomically ubiquitous 51112131415. However, the uterus and retroperitoneum have emerged as the 2 most frequently reported sites of origin for these neoplasms 57812. Notably, most of the uterine cases were described within the past decade 8. The first uterine PEComa was described in 1992 19, and 43 additional cases have subsequently been reported in the English literature 51320212223242526272829303132333435363738394041. In this article, the author evaluates the current state of knowledge on PEComas of the uterus, with an emphasis on controversial areas and their unclear relationship with uterine smooth muscle neoplasia.

Clinical features of the 44 previously reported cases of uterine PEComa are summarized in Table 1. These cases include tumors reported as PEComa, perivascular epithelioid cell tumor, abdominopelvic sarcoma, hyalinized uterine mesenchymal neoplasms with HMB-45-positive epithelioid cells, epithelioid angiomyolipoma 5131920212223242526272829303132333435363738394041, and excludes cases reported as conventional angiomyolipoma, lipoleiomyomas and lymphangioleiomyomatosis of the uterus 4243444546474849. One case that was originally reported as lymphangioleiomyomatosis 50, but which was subsequently included in a series of PEComas 33, is included. Only cases reported prior to January 31, 2008 in the English literature are included.

Forty and four of these 44 cases arose in the uterine corpus and uterine cervix respectively. The patients ranged in age from 9 to 79 (mean 45). The presentations were wide and varied, and included abnormal vaginal bleeding, abdominopelvic pain, uterine rupture, and hemoperitoneum; the tumors were incidental discoveries in several cases 5131920212223242526272829303132333435363738394041. The radiological appearances have been similarly varied and appear to be largely dependent on the biologic characteristics of the underlying tumors. They may be small and homogeneous, simulating a benign smooth muscle neoplasm or they may be large, lobulated and heterogeneous masses 51. Overall, neither the clinical presentation nor radiologic appearance of uterine PEComas is sufficiently distinctive to allow the diagnosis to be suggested preoperatively.

Preliminary data is suggestive of a possible association between uterine PEComa and the tuberous sclerosis complex (TSC). TSC is an autosomal dominant syndrome, one of the phacomatoses, that may be characterized by a wide variety of neoplastic manifestations, including renal angiomyolipomas, lymphangioleiomyomatosis, cardiac rhabdomyomas, subependymal giant cell astrocytoma as well as several others 52. Four (9.1%) of the 44 patients with uterine PEComas also had the TSC 53334. Although this rate of association is probably inflated due to the selection bias associated with the reported cases, it is still notably higher than any other neoplastic process of the uterus, and suggests that patients with uterine PEComa be briefly evaluated for the stigmata of this complex. Lymphangioleiomyomatosis, another characteristic manifestation of TSC, was identified in the lymph nodes of 3 (6.8%) of the 44 patients 3341, 2 of whom had TSC. The vast majority of the 44 patients received surgical management that included at least a hysterectomy.

The reported uterine PEComas have displayed a spectrum of biologic behaviors, and any discussions about the "prognosis" of PEComas as a single neoplasm is fraught with the same fallacies of rendering a blanket statement about the prognosis of a heterogeneous group such as uterine smooth muscle tumors, for example. Follow-up information and/or manifest malignancy at presentation were available in 37 (84%) of the 44 patients. In a previous report 8, a subset of these cases (corpus only) was classified into 2 groups based on patient outcome. The first group was designated "malignant" and was comprised of cases associated with patient death of disease and/or extrauterine extension at presentation. The second group was designated "non-malignant" and was comprised of cases in which neither of the aforementioned features was present 8. After updating that paradigm with cases reported since the report, 15 (44%) of the 34 corpus cases (with follow-up and/or manifest malignancy at presentation) may be classified as malignant and the remaining 19 (56%) as non-malignant. If the 3 cases primary in the cervix and with follow-up information are included, there would be 16 (43%) of 37 cases in the malignant group and 21 (57%) of 37 cases in the non-malignant group. There is no statistically significant difference in patient age between the 2 groups. The above analysis excludes the uterine case included in the series of Pan et al due to lack of specific data 53.

The uterine tumors reported as PEComas have been fundamentally characterized by a diffuse, nested and/or fascicular proliferation of spindled and epithelioid cells that display clear to eosinophilic cytoplasms, and which display immunoreactivity for melanocytic markers 5131920212223242526272829303132333435363738394041. A given lesion may be dominated by spindled cells or epithelioid cells, but these cells are often admixed 813. Some cases are comprised of cells with bland nuclei whereas others display frank anaplasia; most fall somewhere within this spectrum. Multinucleated giant cells, which appear to be degenerative in nature may also be found 581341 (Figure 1). The constituent cells may display a perivascular distribution, but this feature is often inconspicuous in the uterine cases 8. PEComas typically display a prominent network of small capillaries reminiscent of renal clear cell carcinoma or myxoid liposarcoma (Figure 1). The tumors may be circumscribed with no or minor areas of peripheral infiltration, or display a prominent "tongue-like" infiltration into the myometrium 833. Two cases 541 were associated with clusters of epithelioid cells outside of the main tumoral mass, including the ovary and small bowel in one case 5. This phenomenon was designated "PEComatosis" in one 2004 report 5. PEComas may display stromal hyalinization that can be so diffuse as to obfuscate their underlying features 192533353941. In addition to HMB-45, PEComas also display immunoreactivity, albeit in lesser proportions, to other melanocytic markers such as micropthalmia transcription factor, Melan A, Mart-1, and HMSA-1. Greater than 70% are positive for smooth muscle actin (SMA), and almost half are positive for desmin 8. The full immunophenotype of uterine PEComas is outlined in Table 2.

There have been three major attempts to correlate pathologic parameters of PEComas with patient outcome, in an effort to enumerate morphologic criteria predictive of malignant potential 5813. Folpe et al 13 derived their proposed criteria from a review of PEComas reported from a wide variety of anatomic locations. The resultant classification included 3 groups: 1) A "Benign" group in which none of 22 cases displayed aggressive behavior and which was characterized by tumor non-infiltrativeness, tumor size less than 5 cm, non high nuclear grade, lack of high cellularity, mitotic rate ≤ 1/50 high power fields (HPF), no necrosis and no vascular invasion; 2) A "malignant" group in which 12 (71%) of 17 cases displayed aggressive behavior and which was characterized by tumors grater than 5 cm, infiltrativeness, high nuclear grade and cellularity, necrosis, lymphovascular invasion and a mitotic rate ≥ 1/50 HPF; and 3)a group of tumors of "uncertain malignant potential" that were characterized by tumors with nuclear pleomorphism/multinucleated giant cells [0 (0%) of 6 displaying aggressive behavior] or size greater than 5 cm only [2 (12%) of 17 cases displaying aggressive behavior] 13. In our own aforementioned analysis of 31 corpus cases which were classified based on patient outcome into "Non-Malignant" and "Malignant" groups, significant differences were found between these 2 groups regarding 3 pathologic parameters 8. Regarding tumor size, the malignant cases (average size 9.6 cm) were significantly larger than their non-malignant counterparts (average size 4.67 cm, p = 0.04). However, there were no size thresholds that in of themselves could classify even 75% of the cases in both groups. The presence of coagulative necrosis (Figure 2) was highly associated with the malignant group [present in 9 (82%) of 11 cases], as compared with only 2 (11.8%) of the 17 cases in the non-malignant group (p = 0.0002). Finally, a mitotic count of ≤ 1 mitotic figures/10 HPF was found in 16 (88%) of the 18 non-malignant cases but in only 4 (40%) of the 10 malignant cases (p = 0.01). Although the possibility of suboptimal sampling remains, it is noteworthy that some uterine PEComas have been reported in which metastases developed in the absence of mitotic activity 2329. As such, lack of mitotic figures may not be necessarily reassuring for non-malignancy. Application of the Folpe et al 13 criteria to the uterine corpus cases showed that it classified 12 of the 13 malignant cases under our paradigm 8 appropriately into a malignant group. The 13^th case 32 would probably be classified as being of uncertain malignant potential. Because nuclear atypia and nuclear pleomorphism were inconsistently defined in many of the reported cases of uterine PEComa, The Folpe et al 13 criteria would probably also classify most of our "non-malignant" cases into their "uncertain malignant potential" group simply because of nuclear pleomorphism. Furthermore, "infiltrativeness", which presumably encompasses the "tongue-like" myometrial growth pattern in some uterine PEComas, would remove a tumor that otherwise qualifies from the "benign" group in the Folpe et al 13 classification, which is at best questionable regarding correlation with outcome. None of the 3 cases with this infiltrative pattern (and follow-up) in the series of Vang and Kempson 33, for example, recurred. The case associated with intraabdominal PEComatosis reported by Fadare et al 5 was otherwise histologically benign and also did not recur.

As with conventional smooth muscle neoplasms of the uterus, there are no morphologic criteria that can uniformly predict the malignant potential of all cases. Nonetheless, the presence of coagulative necrosis and/or a mitotic index >1 mitotic figure (MF)/10 HPF is highly worrisome for malignancy and patients with such tumors should be managed as such. However, a mitotic index >1 MF/10 HPF is typically seen in association with other features such as necrosis and cytologic atypia. Cases that can be classified as benign according to Folpe et al criteria 13 should be reported as being highly unlikely to display aggressive behavior based on limited evidence. In the author's opinion, until more cases are described and prognostic criteria become more finessed, all patients diagnosed with uterine PEComas should receive long-term follow-up irrespective of the pathologic designation because there is an inherent element of unpredictability to these neoplasms.

The principal controversial aspect of uterine PEComas stems from the fact that they have a distinct clinicomorphologic and immunophenotypic overlap with some smooth muscle neoplasms of the myometrium, which has called into question the validity of their segregation as a distinct clinicopathologic entity. In the simplest analysis of the overlap, smooth muscle neoplasia may be viewed as uterine tumors that always display myomatous differentiation and occasionally display melanocytic differentiation, whereas PEComas always display melanocytic differentiation and usually displays some myomatous differentiation. From this construct, it becomes apparent how controversies may arise regarding which end of this putative spectrum is best considered a variant of the other and which designation is most appropriate to apply for a given case that falls within the areas of overlap. These issues are discussed below.

Given the aforementioned extent of clinicomorphologic and immunophenotypic overlap that exists between uterine PEComas and ESM (Table 3), and since 42 (95%) of the 44 uterine PEComas were reported since the beginning of 2000, the logical question arises as to whether PEComas are a distinct clinicopathologic entity or whether they merely represent a selected group of HMB-45 positive smooth muscle tumors. In the author's opinion, this is a false choice. First, the discussion has to be "purified" by the stipulation that a uterine mesenchymal tumor should not be designated a PEComa simply because it is positive for HMB-45 or some other melanocytic marker. There is no significant debate at this junction that conventional smooth muscle tumors (i.e. tumors that are comprised predominantly of fascicles of spindle cells with eosinophilic cytoplasm) can display immunoreactivity for melanocytic markers 6465. When this group of tumors is removed the discussion can then be centered on uterine mesenchymal tumors that are comprised predominantly of epithelioid cells or an admixture of epithelioid and spindle cells throughout the tumor.

A clinicopathologic entity should be pathologically (via morphologic evaluation and/or ancillary techniques) definable and have clinical significance. ESM have been so classified for at least three decades. However, their behavior has been notoriously difficult to predict from morphologic criteria, which may be considered as evidence of excessive heterogeneity in tumors currently defined as ESM and the need to delineate biologically relevant subsets within them. PEComas and ESM have significant similarities but also significant differences (Table 3). Indeed, an argument can be advanced that it is no more invalid a position to consider PEComas as a variant of ESM than it is to consider all previously reported ESM as variants of PEComas. What is undisputable is that these tumors share lines of differentiation. It is an impediment to scientific progress when scientists take rigid, dogmatic positions simply because of tradition. One approach to resolving the PEComa versus ESM question is to postulate that a given tumor is better characterized into a diagnostic category if pathologically similar, extrauterine tumors are well described. In this respect, it is noteworthy that epithelioid smooth muscle tumors are decidedly rare outside of the uterus. Most of the extrauterine (predominantly retroperitoneal and gastrointestinal) tumors that were previously diagnosed as epithelioid smooth muscle tumors are now considered epithelioid gastrointestinal stromal tumors. In contrast, PEComas are considered anatomically ubiquitous. One comparative genomic hybridization study found a closer kinship, regarding the patterns of chromosome losses and gains, between a uterine PEComa and extrauterine PEComas than between the uterine PEComa and uterine smooth muscle neoplasia 53. In animal models carrying a germline mutation to the tuberous sclerosis 2 (TSC2) gene, there is an increased predisposition to develop uterine and extrauterine tumors 79. Notably, a disproportionate percentage of the uterine tumors, which the authors classified as leiomyomas/leiomyosarcomas based on morphology and desmin/actin immunoreactivity, were of the epithelioid type. PEComas, of course, have a known association with TSC2 gene alterations 80. Second, a recent study reported a 100% rate of CD1a immunoexpression in PEComas from various sites 81. The author examined 5 ESM (4 epithelioid leiomyomas and 1 epithelioid leiomyosarcoma) diagnosed by WHO criteria 17 and all were negative for CD1a. (Fadare O, unpublished data, 2008). Other selectively noteworthy findings in uterine PEComas include their 9.1% association with TSC, their occasional occurrence in young patients 232934, and their occasional display of metastases in the absence of significant tumor mitotic activity 2329. Although the precise etiopathogenesis of PEComas remain to be elucidated, the findings outlined above argue for PEComas as a distinct entity or at least a specific subset of smooth muscle neoplasia. The recent demonstration of elevated phospho-p70S6K and reduced phospho-AKT expression, indicative of mTOR (mammalian target of rapamycin) activation, in a group of extra-renal PEComas raises the possibility of using mTOR inhibitors such as rapamycin in the treatment of uterine PEComas and is another argument for their routine segregation 82.

The author shares the opinion that has been expressed by others that most PEComas can be morphologically distinguished from classical epithelioid smooth muscle tumors by their distinctive network of capillaries 13. Nonetheless, these lesions may exist at different points on a single clinicomorphologic spectrum 33, and their distinction may admittedly be difficult. Future studies should evaluate a series of archival epithelioid smooth muscle tumors to determine whether cases that are morphologically and immunophenotypically more consistent with PEComas are identifiable, and perhaps more importantly, whether these cases are prognostically distinct from the remaining cases in the group. At present time, the author proposes this practical approach to these neoplasms is as follows:

1) Predominantly conventional smooth muscle neoplasms that display immunoreactivity for one or more melanocytic markers should be diagnosed as leiomyosarcomas/leiomyomas, with a comment about their melanocytic differention. This is designed to segregate these cases because the significance of melanocytic differentiation is unknown, and melanocytic differentiation may emerge at a metastatic site 59, with the attendant potential for the misdiagnosis of the metastatic lesion as a primarily melanocytic malignancy. The presence of clear or epithelioid cells in an otherwise predominantly conventional smooth muscle neoplasm should not affect its designation as such.

2) All epithelioid mesenchymal neoplasms of the uterus should be tested for melanocytic markers, to include at least HMB-45.

3) Tumors with absolutely characteristic morphologic and immunophenotypic features are diagnosed as a perivascular epithelioid cell tumor (PEComa). However, given the current state of evidence, it is no longer advisable to use "PEComa" as a diagnostic term in isolation, as it is akin to diagnosing a uterine mesenchymal tumor as a "smooth muscle tumor" without further description. Rather, there should be some statement regarding their expected biologic behavior as predicted by morphologic characteristics. PEComas should be described as being highly likely to display aggressive behavior if they display coagulative necrosis. A mitotic index of >1/10 HPF is highly worrisome for malignancy but have been uniformly seen concurrent with other features such as cytologic atypia and necrosis. PEComa cases that can be classified as benign according to Folpe et al criteria 13 should be reported as being highly unlikely to display aggressive behavior based on limited evidence. All other cases (PEComas) are of uncertain malignant potential. It is recommended that all patients diagnosed with uterine PEComas receive long-term follow-up irrespective of the pathologic designation due to the unpredictability to these neoplasms.

4) Epithelioid mesenchymal tumors that do not display all the characteristic features of PEComa should be diagnosed as epithelioid leiomyomas, epithelioid leiomyosarcomas or epithelioid tumors of uncertain malignant potential (if they otherwise have the features of these lesions) per aforementioned criteria 175455565761. As with their conventional counterparts, melanocytic differentiation should similarly be noted if present.

5) Finally, as with all neoplasms, adequate sampling is of paramount importance to ensure that all areas of these uterine mesenchymal tumors are being optimally represented microscopically.

The author(s) declare that they have no competing interests.

OF wrote the manuscript.